Serverless Agents: Real-World Tooling with Strands SDK, MCP & AWS
summary
Demonstrates a concrete pipeline: MCP JSON-RPC describes tools so any LLM discovers and calls them, Strands SDK provides Strands SDK agentic tools and model-driven orchestration, and AWS Lambda hosts stateless MCP servers with CDK deployment and optional layers to avoid bloat. Shows a weather tool and a database agent that runs natural‑language queries against Postgres and returns a 30‑day revenue summary with peak days (demo uses screenshots due to flaky internet). Emphasizes parameterized queries, secrets manager, minimal IAM, and horizontal scaling.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:34Problem with traditional AI agents
- 1:53Model Context Protocol (MCP)
- 6:23Strands SDK
- 7:26Example: Weather tool
- 10:22Example: Database agent
- 12:42Example: Tool definition
- 15:25Patterns & best practices
- 17:11Security
- 18:51Scaling & evolution
- 19:37Key takeaways
- 20:52Outro
related talks
19mBuilding Agents at Cloud Scale
16mBuilding Agentic AI applications with Positron and AWS Strands Agents
29mBuilding Production-Ready AI Agents
14mIntroducing Strands Agents, an Open Source AI Agents SDK
28mBuilding Agentic AI applications on AWS
06mKeynote: Building Agents at Scale
21mBuilding Agents (the hard parts!)
36mPutting AI Agents to Work: A New Era of Open Source Connectivity with MCP